Cyclic di-GMP, an established secondary messenger still speeding up

Authors


E-mail ute.romling@ki.se; Tel. (+46) 8 524 87319; Fax (+46) 8 330744.

Summary

The secondary messenger cyclic di-GMP coordinately regulates the transition between motility/sessility/virulence in bacterial populations and upon adaptation to novel habitats. Thereby, multiple independent regulatory circuits regulate a diversity of targets. This specific output response is surprising considering the diverse physiological processes regulated by this signalling molecule, which range from transcription to proteolysis and clearly demonstrates the presence of sophisticated developmental programmes in these so-called simple organisms.

Introduction

The career of the small molecule cyclic di-GMP is still speeding up, although starting more than 20 years ago. Cyclic-di-GMP was identified as an allosteric activator of the cellulose synthase in the fruit rotting bacterium Gluconacetobacter xylinus (Ross et al., 1987). Successively, as cyclic di-GMP could not fulfil the expectation to be an activator of cellulose biosynthesis in plants (Delmer, 1999), there was a long lag-phase in the research about this molecule. However, the identification of proteins that ‘make and break’ cyclic di-GMP, di-guanylate cyclases and cyclic di-GMP specific phosphodiesterases (Tal et al., 1998), together with the identification of protein domains with their potential catalytic activity in almost any bacterial genome, laid the ground for the 2004 experimental demonstration that cyclic di-GMP is a global secondary messenger exclusively found in bacteria (Paul et al., 2004; Simm et al., 2004; Tischler and Camilli, 2004). Transition between fundamental bacterial behaviour such as sessility, motility and virulence is regulated by the activity of GGDEF domain di-guanylate cyclases and EAL and HD-GYP domain phosphodiesterases (Fig. 1; D'Argenio and Miller, 2004; Jenal, 2004; Römling et al., 2005; Ryan et al., 2006; Cotter and Stibitz, 2007). What have we learnt since then?

Figure 1.

Components and responses of cyclic di-GMP signalling pathways. GGDEF domains proteins synthesize cyclic di-GMP, while EAL and HD-GYP domain proteins degrade cyclic di-GMP. Cyclic di-GMP signalling affects fundamental physiological processes such as transcription, translation, post-translational events and proteolysis. Protein and RNA-based cyclic di-GMP receptors which mediate these responses have been identified. Transcription is influenced by cyclic di-GMP binding to the transcriptional regulators Clp, FleQ and VspT (Hickman and Harwood, 2008; Chin et al., 2009; Leduc and Roberts, 2009; Krasteva et al., 2010; Tao et al., 2010). Binding of cyclic di-GMP to the riboswitche classes c-di-GMP-I and c-di-GMP-II affects transcriptional elongation and translation (Sudarsan et al., 2008; Lee et al., 2010b). RNA processing and degradation is altered when cyclic di-GMP binds to polynucleotide phosphorylase (PNPase) (Tuckerman et al., 2011), which is a component of protein complexes dedicated to RNA degradation. Binding of cyclic di-GMP to enzymatically inactive GGDEF and EAL domains with degenerated catalytic motifs as present in PopA and LapD, respectively, alters proteolysis through protein–protein interactions (Duerig et al., 2009; Newell et al., 2011). Proteolysis of the surface protein LapA is inhibited by binding of cyclic di-GMP to LapD. Binding of cyclic di-GMP to PilZ domain proteins alters protein-protein interactions. The flagella rotation speed and chemotaxis is affected through binding of cyclic di-GMP to the PilZ domain protein YcgR to flagella motor components (Boehm et al., 2010; Fang and Gomelsky, 2010; Paul et al., 2010). In addition, the biosynthesis of exopolysaccharides such as cellulose is stimulated by cyclic di-GMP signalling (Ross et al., 1987).

Despite all – bacteria can live without it!

Although often stated to be ubiquitous, bacteria can certainly live without cyclic di-GMP. Indeed, bacteria can successfully form biofilms and be virulent without cyclic di-GMP signalling pathways (Holland et al., 2008). The majority of bacterial species with a genome below 2 Mbp and even 15% of bacterial species with a genome over 2 Mbp do not show bioinformatic indications of cyclic di-GMP production (Seshasayee et al., 2010). Do those bacteria not need to switch between sessility and motility, probably the most fundamental lifestyle change controlled by cyclic di-GMP? Does the lifestyle of those bacteria not require a down to millisecond-fast cellular response with high amplification to extra- and intracellular signals to rapidly adapt physiological processes? Or has another secondary signalling molecule such as cAMP, cGMP or even c-di-AMP (Römling, 2008; Witte et al., 2008; Gomelsky, 2011; Marden et al., 2011) taken over a similar function?

Essentially, the complexity of cyclic di-GMP signalling systems is not linearly correlated with genome size, indicating that this signalling system is highly flexible. The highly modular nature of this signalling system is also indicated by the gene arrangement, as cyclic di-GMP metabolizing proteins are usually encoded by stand-alone genes. Indeed, related genomes can have highly variable number of cyclic di-GMP metabolizing proteins and cyclic di-GMP signalling can even be readily eliminated in closely related bacteria. For example, species within the genus Clostridium can have between 0 and 43 potential cyclic di-GMP metabolizing proteins and in species of the genus Mycobacterium the number of these proteins can range from 0 to 22 (Galperin et al., 2010; Bordeleau et al., 2011). Gradual loss of complexity in cyclic di-GMP signalling can be observed between recently evolved species. After its divergence from Yersinia pseudotuberculosis 20 000 years ago accompanied with habitat restrictionand increased human virulence, Yersinia pestis has maintained only three of six cyclic di-GMP metabolizing proteins functional (Bobrov et al., 2011). Thus, although lifestyle rather than phylogeny seems to determine the complexity of cyclic di-GMP signalling, some evolutionary forces still remain obscure. For example, Proteus mirabilis, a species widely distributed in soil and water and at the same time an urinary tract pathogen has only one potential di-guanylate cyclase GGDEF protein (Pearson et al., 2008), whereas Escherichia coli, another member of the family Enterobacteriaceae with a similar lifestyle, has 31 cyclic di-GMP metabolizing proteins (Galperin et al., 2010).

Biofilm formation requires complex cyclic di-GMP signalling networks

Cyclic-di-GMP signalling promotes biofilm formation in many bacteria. Usually, in one bacterial strain a variety of cyclic di-GMP metabolizing proteins regulate biofilm formation (Garcia et al., 2004; Kulesekara et al., 2006; Lim et al., 2006; Simm et al., 2007; Sommerfeldt et al., 2009). At the first glance, it may seem surprising that a single phenotype (biofilm formation) is subject to such a sophisticated regulation. And there is still a debate going on whether simple bacteria have dedicated developmental programs to regulate biofilm formation and are not simply responding more or less spontaneously to changing environmental conditions (Ghigo, 2003; Monds and O'Toole, 2009). In any case, there is an agreement that biofilm formation is a highly complex process where the expression of a variety of extracellular matrix components is temporally and spatially regulated to develop a sophisticated three-dimensional biofilm architecture.

One can take the example of Pseudomonas aeruginosa where biofilm formation in its complexity has been thoroughly studied. The exopolysaccharide alginate, a variety of other exopolysaccharides, nucleic acid and proteins, such as cup fimbriae and type IV pili, all contribute to the formation of the biofilm extracellular matrix and thus built up multicellular bacterial aggregates (O'Toole and Kolter, 1998; Vallet et al., 2001; Whitchurch et al., 2002; Ma et al., 2009; Borlee et al., 2010). Emerging cumulative data indicate that the expression and/or biosynthesis of many of these biofilm matrix components are regulated by specific cyclic di-GMP signalling networks, which act on different levels and can partially overlap (Lee et al., 2007; Merighi et al., 2007; Mikkelsen et al., 2011). For example, expression of the Pel and Psl polysaccharides requires the di-guanylate cyclase YfiN (Ueda and Wood, 2009; Malone et al., 2010), while in some strains alginate biosynthesis is activated by PA1727 (Hay et al., 2009). On the other hand, the expression of CupA fimbriae is activated by the putative di-guanylate cyclase SiaD (PA0169) upon exposure of P. aeruginosa to the detergent sodium-dodecylsulfate, which leads to cell aggregation (Klebensberger et al., 2009). Besides transcriptional regulation, the Pel and alginate exopolysaccharides are regulated by cyclic di-GMP also at the post-transcriptional level through cyclic di-GMP receptors associated with the macromolecular biosynthesis complex (Lee et al., 2007; Merighi et al., 2007). Although our knowledge about the cyclic di-GMP network components which affect the synthesis, degradation of cyclic di-GMP as well as the cyclic di-GMP receptor(s) which regulate one target output is still fragmentary, one would expect distinct di-guanylate cyclases, phosphodiesterases and receptors to adjust the cyclic di-GMP signal on the different regulatory levels.

Cyclic di-GMP signalling regulates biofilm formation on the single cell level

Biofilms are so much more than just cell aggregates. They possess a sophisticated three-dimensional architecture often referred to as the ‘mushroom structure’, which is required for full expression of biofilm features such as elevated antimicrobial resistance (Davies et al., 1998; Bjarnsholt et al., 2005; Bridier et al., 2011). This structure is essentially a balance between multicellular complexes of bacterial cells and open spaces. Consequently, besides by adhesive extracellular matrix components, which promote cell mass accumulation and sessility, a mature biofilm architecture is shaped by motile cells, expression of anti-adhesive molecules, cell lysis and biofilm dissolution (Klausen et al., 2003; Barraud et al., 2009; Ueda and Wood, 2010). While it has been well established that motility is stimulated by the absence of cyclic di-GMP signalling (Wolfe and Visick, 2008), it has recently been shown that in P. aeruginosa cell lysis and extracellular DNA release are stimulated by low cyclic di-GMP concentrations and inversely associated with biofilm formation (Ueda and Wood, 2010). This complex regulation of biofilm formation has consequences on how cyclic di-GMP signalling networks are regulated to build up mature mushroom-like biofilm structures. In conclusion, cyclic di-GMP signalling must determine the fate of bacteria at the single cell level as previously predicted (Simm et al., 2004). The bistable expression, a situation where cells in a population show either ON or OFF expression, of the major biofilm regulator CsgD in S. typhimurium biofilms provides indirect evidence for a differential regulation of protein expression by cyclic di-GMP signalling on the single cell level. CsgD expression is positively regulated by cyclic di-GMP signalling on the transcriptional and post-transcriptional level (Kader et al., 2006; Simm et al., 2007). A rise in cyclic di-GMP concentration caused by inactivation of a cyclic di-GMP phsophodiesterase, however, lead to high CsgD expression in all cells (Grantcharova et al., 2010), suggesting that differences in the local or overall cyclic di-GMP concentration in individual cells are one factor which determines the bistable expression of CsgD. But is there evidence for asymmetrical distribution of cyclic di-GMP in cells? Indeed, recently, the asymmetrical distribution of cyclic di-GMP in dividing cells could be directly monitored with the cyclic di-GMP receptor YcgR (Christen et al., 2010). Cyclic di-GMP concentrations were found to be coherent with bacterial behaviour whereby the flagella-bearing motile cell had lower concentrations of cyclic di-GMP, while the non-flagellated cell had higher concentrations of cyclic di-GMP. How can such a different distribution of cyclic di-GMP which is basically a freely diffusible molecule be achieved in individual cells? Spatial distribution of di-guanylate cyclases, phosphodiesterases and/or cyclic di-GMP receptors with different affinity prior to cell division are possible explanations. But we still have much to learn how cyclic di-GMP signalling contributes to the formation of mature biofilm structures.

Virulence

Although the first cyclic di-GMP dependent processes more closely examined were associated with bacterial development such as biofilm formation and motility, regulation of virulence phenotypes is also a prominent feature of cyclic di-GMP signalling. Cyclic di-GMP signalling can affect virtually almost all kind of virulence phenotypes including the overall in vivo virulence phenotype in animals and plants (Kulesekara et al., 2006; Ryan et al., 2007; Tamayo et al., 2007). Individual phenotypes, for example, interactions with host cells such as adherence to host cells, host cell invasion, cytotoxicity, intracellular infection, secretion of virulence factors and stimulation of immune response have been reported to be affected by cyclic di-GMP signalling (Hisert et al., 2005; Tischler and Camilli, 2005; Kulesekara et al., 2006; Lai et al., 2008; McWhirter et al., 2009; Kumagai et al., 2010; Lee et al., 2010a; Lamprokostopoulou et al., 2010; Sauer et al., 2011).

Based on early investigations, the common view is that an acute infection process does not require cyclic di-GMP (Tamayo et al., 2007). Studies of pathogenesis in pathogens with few cyclic di-GMP metabolizing proteins seem to support this view. Cyclic di-GMP signalling is not required for virulence of Y. pestis in a plague mouse model (Bobrov et al., 2011) and also the Lyme disease spirochaete Borrelia burgdorferi can successfully cause infection in mice without using its cyclic di-GMP signalling system (He et al., 2011). Even more, uncontrolled high cyclic di-GMP levels achieved by deletion of the single phosphodiesterase inhibit acute infection through the expression of extracellular biofilm matrix components (Bobrov et al., 2011; Sultan et al., 2011).

On the other hand, biofilm formation, which is stimulated by cyclic di-GMP signalling, is a virulence factor in chronic infections (Tamayo et al., 2007). Occurrence of small colony variants of P. aeruginosa with elevated cyclic di-GMP levels after year-long persistence in cystic fibrosis lung infection supports this view (Smith et al., 2006; Meissner et al., 2007; Starkey et al., 2009). Those variants showed elevated biofilm formation, enhanced fimbrial expression, repression of flagellar expression and enhanced antibiotic resistance (Drenkard and Ausubel, 2002; Meissner et al., 2007; Starkey et al., 2009). A direct role for cyclic di-GMP signalling in enhanced persistence of P. aeruginosa has recently been demonstrated in a chinchilla middle ear infection model (Byrd et al., 2011). Also bacterial aggregation in host cells, which closely resembles biofilm formation, is mediated by cyclic di-GMP signalling (Kumagai et al., 2011). Thus, in bacterial host interactions cyclic di-GMP signalling seems to regulate the transition between acute and chronic infection, virulence and persistence and maybe even between virulence and commensalism.

Despite these relatively clear-cut regulatory patterns proposed for cyclic di-GMP signalling in acute and chronic infection processes, the role of cyclic di-GMP in the in vivo virulence phenotype in animals and plants, but even in a defined event such as cytotoxicity or secretion of a virulence factor, is elusive. Genetic screens in pathogens with several cyclic di-GMP metabolizing proteins have observed the regulation of a virulence phenotype by cyclic di-GMP metabolizing proteins with opposite functions operating in the same direction (Kulesekara et al., 2006; Ryan et al., 2007; I. Ahmad, A. Lamprokostopoulou, S. Le Guyon, E. Streck, M. Barthel, V. Peters, W.-D. Hardt and U. Römling, unpubl. data). A trivial explanation would be that the proteins do not perform the function that is predicted by bioinformatic analysis. However, several alternative, not necessarily contradictory explanations supported by experimental evidence, can explain this apparent discrepancy.

First, secretion of a virulence factor, an effector protein or an immunostimulatory molecule is already a complex physiological event, which includes the expression of a functional secretion apparatus and the expression and secretion of effector proteins. Imagine that cyclic di-GMP signalling network components with opposite functions are required at different stages of the overall secretion process. For example, the type III secretion system mediates adhesion to host cells and subsequently translocates effector proteins into host cells (Misselwitz et al., 2010). Thus, in line with the sessility-virulence switch hypothesis, it would be expected that the expression of a functional secretion apparatus requires cyclic di-GMP signalling, while the expression and secretion of effector proteins is effective when cyclic di-GMP is lacking.

Second, a certain macromolecular structure can function as an adhesin and in secretion. For example, we have observed that high cyclic di-GMP levels promote the presence of cell-associated flagellin most likely in the form of flagella, but inhibit secretion of monomeric immunostimmulatory flagellin into the culture supernatant (Lamprokostopoulou et al., 2010). This finding is in line with observations that assembled flagella on the bacterial surface are required for biofilm formation on cholesterol gall stone surfaces (Crawford et al., 2010), which promotes the chronic carrier status of Salmonella, while monomeric secreted flagellin is an immunostimulator in acute infection.

Third, on the molecular level, cyclic di-GMP metabolizing proteins of opposing function can interact with each other (Andrade et al., 2006; Ryan et al., 2010). Although the direct molecular consequences of such interactions are not known, deletion of two di-guanylate cyclase GGDEF proteins, which interact with the phosphodiesterase HD-GYP protein RpfG, resulted in a reduction of Xanthomonas campestris pili-mediated motility although one would expect the opposite phenotype (Ryan et al., 2010). As the phosphodiesterase activity of RpfG is required for motility, it is not simply a switch in cyclic di-GMP mediated regulation of pili motility which is occurring.

Forth, investigations of the V. cholerae infection process provided experimental evidence for a temporal and spatial requirement of opposite cyclic di-GMP metabolizing processes at different stages of the disease. In early infection upon entry into the host intestine V. cholerae expresses the cholera toxin under low concentrations of cyclic di-GMP provided by the cyclic di-GMP phosphodiesterase VieA (Tischler and Camilli, 2005). After colonization and proliferation, in the late stage of infection, a group of RpoS regulated GGDEF/EAL genes are induced suggesting increased cyclic di-GMP concentration, but mutants do not have a growth disadvantage in the small intestine of mice (Schild et al., 2007; Tamayo et al., 2008). Instead, the bacteria prepare for their further transition through the host (without disease phenotype), as a mutant in three GGDEF genes is defective in survival in the watery stool from the large bowel (Schild et al., 2007). Thus, if monitoring stool samples as virulence read-out, one would probably find these GGDEF mutants in low concentrations and wrongly conclude that GGDEF domain proteins and cyclic di-GMP signalling have a role in virulence.

Novel phenotypes – a role for cyclic di-GMP in transmission

The life cycle of most human pathogens includes an outside-host phase. For example, Y. pestis and B. burgdorferi are transmitted by arthropod vectors. Although cyclic di-GMP signalling has no role in virulence in Y. pestis and B. burgdorferi, it plays a determinative role in transmission between the arthropods and the human host. For Y. pestis, transmission from the flea tothe human host requires biofilm formation mediated by the poly-β-1,6-N-acetylglucosamine exopolysaccharide (Jarrett et al., 2004), which is positively regulated by cyclic di-GMP signalling. Interestingly, stimulation of biofilm formation in the flea requires a di-guanylate cyclase with only a minor in vitro role in the stimulation of biofilm formation (Sun et al., 2011). B. burgdorferi growth in the tick vector requires its sole di-guanylate cyclase which activates the glycerol transport/catabolic operon (He et al., 2011). As glycerol is produced by ticks as an anti-freezing molecule, activation of glycerol catabolism by cyclic di-GMP signalling contributes to growth of B. burgdorferi the reduced genome of which lacks many metabolic pathways.

Once the intestinal pathogen V. cholerae is being shed from its host it will meet the aquatic environment, its natural habitat. Although without virulence defect, a mutant in three GGDEF domain proteins, which are already expressed in late infection, is defective in survival in pond water (Schild et al., 2007). These three examples clearly show the role of cyclic di-GMP signalling in the adaptation to different environments.

Among other novel phenotypes mediated by cyclic di-GMP signalling are regulation of developmental processes such as aerial-mycelium formation in Streptomyces coelicolor (Tran et al., 2011) and production of antibiotics (Fineran et al., 2007; Tran et al., 2011). As the expression of cyclic di-GMP metabolizing proteins and cyclic di-GMP-dependent processes is specifically stimulated by plant exudates (Ausmees et al., 1999; Matilla et al., 2011), the regulatory role of cyclic di-GMP signalling networks in bacterial-plant interactions is likely to expand (Zhang, 2010).

Emergence of secretion systems as novel targets of cyclic di-GMP signalling

The expression of toxins, secretion of virulence factors, the expression of type III secretion system, secretion of type III secretion system effector components, regulation of adhesins and secretion of flagellin are some of the virulence phenotypes found to be regulated by cyclic di-GMP in plant and animal pathogens (Tischler and Camilli, 2005; Ryan et al., 2007; Tamayo et al., 2007; Ryan et al., 2009; Lamprokostopoulou et al., 2010; Lee et al., 2010a; Yi et al., 2010). On the other hand, secretion or surface presentation of adhesins can contribute to biofilm formation and is also regulated by cyclic di-GMP signalling (Monds et al., 2007; Borlee et al., 2010). These findings point out the secretion systems as novel general targets of cyclic di-GMP signalling (Fig. 2), which can be positively or negatively regulated by cyclic di-GMP signalling. Indeed, all types of secretion systems, from type 1 until type 6, have been shown to be targets of cyclic di-GMP signalling, although in most systems the exact mechanism of cyclic di-GMP regulation is still unknown. As in the case of regulation of flagella biosynthesis and motility (Wolfe and Visick, 2008), the regulation can take place on different levels and include transcriptional regulation of structural components and secreted proteins as well as the direct processing of the protein upon secretion. Low cyclic di-GMP levels activate the expression of the type III secretion system of D. dadantii required for plant virulence through the nitrogen response sigma factor RpoN (Yi et al., 2010). In biofilm formation, the CdrAB two-partner secretion system of P. aeruginosa, which codes for a large non-fimbrial β-helical adhesin is positively regulated by cyclic di-GMP signalling on the transcriptional level (Borlee et al., 2010). On the other hand, expression and secretion of the type 1 secretion substrate LapA protein is affected by cyclic di-GMP signalling upon overexpression of a cyclic di-GMP phosphodiesterase (Monds et al., 2007). However, the transition between a secreted adhesive form of the protein and a proteolytically processed supernatant-released form of the protein is regulated by a specific cyclic di-GMP signalling network responding to the environmental levels of inorganic phosphate (Newell et al., 2011).

Figure 2.

Secretion systems as targets of cyclic di-GMP signalling. A. The type 1 secretion system (T1SS) uses ATP-binding cassette (ABC) transporters to secrete proteinaceous and non-proteinaceous substrates. The LapA protein of Pseudomonas fluorescence has been shown to be processed upon secretion by the periplasmic proteolytic activity of LapG which is inhibited by cyclic di-GMP signalling (Newell et al., 2011). The proteolysis regulates transition between the surface-attached and supernatant-secreted form of the protein. The site of action of the cyclic di-GMP affected process is indicated by a red arrow. B. The type two secretion system (T2SS) secrete enzymes and toxins across the outer membrane of Gram-negative bacteria, which are transported by the general secretion pathway (Sec) or the Tat pathway across the inner membrane. Secretion of T2SS substrates has been shown to be affected by cyclic di-GMP signalling in several animal and plant pathogenic bacteria (Tischler and Camilli, 2005; Ryan et al., 2007; Yi et al., 2010). In V. cholerae, expression of the T2SS substrate cholera toxin (CT) is inhibited by cyclic di-GMP signalling on the transcriptional level through inhibition of the transcription of the transcriptional activator ToxT as indicated by the red arrow. C. The two-partner system (Type five b secretion system (T5bSS)) consists of an effector protein and an outer membrane protein transporter. The expression of the non-fimbrial β-helix adhesin CdrA of P. aeruginosa belonging to the CdrA/CdrB two partner system is positively regulated by cyclic di-GMP signalling on the transcriptional level as indicated by the red arrow (Borlee et al., 2010). D. The complex molecular structure of the type three secretion system (T3SS) translocates effectors directly to the cytoplasm of host cells. Expression of the secretion apparatus and secretion of effector proteins has been shown to be negatively regulated by cyclic di-GMP signalling in plant and animal pathogenic bacteria (Lamprokostopoulou et al., 2010; Yi et al., 2010). Cyclic di-GMP signalling regulates expression of the T3SS secretion system in D. dadantii through the repression of the sigma factor RpoN as indicated by the red arrow. E. The type four secretion system (T4SS) translocates proteins and single-stranded DNA. Expression of T4SS components has been shown to be affected by cyclic di-GMP signalling (Kumagai et al., 2010). The regulatory mechanisms of cyclic di-GMP signalling affecting T4SS secretion are unknown. F. The type six secretion system (T6SS) is widespread in Gram-negative bacteria. In P. aeruginosa, expression of Hcp1, the main component of the T6SS nanotube, is positively affected by cyclic di-GMP signalling (Moscoso et al., 2011). The regulatory mechanisms of cyclic di-GMP signalling affecting T6SS expression are unknown.

Coordinated regulation of the motility-sessility and acute-chronic infection switch by cyclic di-GMP signalling

Transition from one lifestyle to the other requires coordination of more than one feature. Thus, the finding that cyclic di-GMP metabolizing proteins are regulated in groups (Bruggemann et al., 2006) is a logical consequence. Also the action of di-guanylate cyclases needs to be counteracted by phosphodiesterases to ensure a restricted physiological response (Gualdi et al., 2007).

Global regulators such as the carbon storage regulator CsrA in E. coli and Salmonella typhimurium and the quorum sensing regulator HapR in Vibrio cholerae coordinate this ‘group behaviour’ of cyclic di-GMP metabolizing proteins directly on the transcriptional and post-transcriptional level (Weber et al., 2006; Schild et al., 2007; Fong and Yildiz, 2008; Hammer and Bassler, 2008; Jonas et al., 2008; Waters et al., 2008; Jonas et al., 2009; Jonas et al., 2010), and thus are major regulators of motility versus sessility or biofilm formation versus virulence.

The coordinated response of cyclic di-GMP networks not only enables fine-tuned output responses on the single cell level in bacterial populations, but also dramatic switches when adaptation to a different habitat is required. For example, CsrA in S. typhimurium strongly activates the phosphodiesterase YhjH and suppresses seven GGDEF/EAL domain proteins directly and indirectly (Jonas et al., 2010). As the CsrA-activated phosphodiesterase YhjH stimulates motility and invasion of host epithelial cells by S. typhimurium, while the CsrA-suppressed cyclic di-GMP metabolizing proteins inhibit motility and/or invasion (Simm et al., 2007; Jonas et al., 2010; I. Ahmad, A. Lamprokostopoulou, S. Le Guyon, E. Streck, M. Barthel, V. Peters, W.-D. Hardt and U. Römling, unpubl. data), CsrA is consequently a major regulator of sessility-motility transition and suggested to control the switch between different physiological states in the infection process (Lucchetti-Miganeh et al., 2008).

Molecular mechanisms of cyclic di-GMP signalling

Identification of cyclic di-GMP receptors is essential for the elucidation of the mechanisms of cyclic di-GMP signalling. Identification of cyclic di-GMP receptors showed that cyclic di-GMP signalling does not only regulate the enzymatic activity of exopolysaccharide synthases (Weinhouse et al., 1997; Lee et al., 2007; Merighi et al., 2007), but affects basal physiological processes such as transcription, RNA processing and degradation, translation and proteolysis (Fig. 1; Hickman and Harwood, 2008; Sudarsan et al., 2008; Lee et al., 2010b; Newell et al., 2011; Tuckerman et al., 2011). Cyclic di-GMP receptors respond to cyclic di-GMP concentrations in the nM to µM range (Sudarsan et al., 2008; Hengge, 2009).

The three transcription factors, Clp, FleQ and VpsT, are not homologous at the amino acid level and differ in their domain structure. The activity of these three factors is altered upon cyclic di-GMP binding with different outcomes. The global regulator Clp (cAMP receptor like protein) and the multidomain σ54 dependent transcription factor FleQ intrinsically bind DNA with high affinity and become inactive upon cyclic di-GMP binding (Hickman and Harwood, 2008; Chin et al., 2009; Leduc and Roberts, 2009; Tao et al., 2010). Thereby, Clp acts mainly as an activator and the inactive Clp-cyclic di-GMP complex subsequently leads to downregulation of the expression of virulence factors in Xanthomonas spp. On the other hand, FleQ works as a repressor of the pel exopolysaccharide operon in P. aeruginosa, the inactive cyclic di-GMP–FleQ complex leads to the relief of repression, subsequent polysaccharide production and biofilm formation. In contrast to Clp and FleQ, cyclic di-GMP binding to the response regulator VpsT activates this transcription factor and leads to the expression of the vps exopolysaccharide in V. cholerae (Krasteva et al., 2010).

Clp, FleQ and VpsT are members of large protein families. Clp is a member of the CRP/FNR superfamily with the cAMP receptor CRP of E. coli as the best-studied member. Surprisingly, docking studies predict that cyclic di-GMP does not bind to an adapted equivalent of the cAMP binding pocket in the cNMP domain, but in the hinge region between the N-terminal cNMP and C-terminal HTH (helix–turn–helix) DNA binding domain (Chin et al., 2009). Binding of cyclic di-GMP to the CsgD-like response regulator VpsT drives protein dimerization, which subsequently leads to DNA binding and activation of exopolysaccharide genes involved in biofilm formation (Krasteva et al., 2010). The four-residue-long conserved W[F/L/M][T/S]R binding motif for cyclic di-GMP is, however, not conserved in the VpsT orthologue CsgD of S. typhimurium and E. coli, which also does not bind cyclic di-GMP (Zakikhany et al., 2010). Altogether, the data show that cyclic di-GMP binding sites can evolve rapidly even in closely related proteins.

Another class of newly discovered cyclic di-GMP receptors are riboswitches, complex RNA structures with the capability to bind to and respond to small metabolites. Up to now, two classes of cyclic di-GMP responsive riboswitches, c-di-GMP-I and c-di-GMP-II, have been discovered present in a wide variety of bacterial species (Sudarsan et al., 2008; Smith et al., 2009; Lee et al., 2010b). These riboswitches, which bind cyclic di-GMP with surprisingly high affinity, are usually located in the 5′ untranslated region of genes and were demonstrated to affect transcriptional elongation and translation.

The polynucleotide phosphorylase (PNPase) of E. coli is another recently identified cyclic di-GMP target (Tuckerman et al., 2011). PNPase is involved in RNA processing and catalyses, among other reactions, the addition of polyA tails to RNA molecules. This finding couples cyclic di-GMP signalling to the regulation of RNA processing and degradation in bacteria.

Predicted by bioinformatic analysis to be a cyclic di-GMP receptor (Amikam and Galperin, 2006), the PilZ domain was the first cyclic di-GMP binding protein to be experimentally identified (Ryjenkov et al., 2006; Christen et al., 2007; Pratt et al., 2007). PilZ domains are short protein stretches of approximately 120 amino acids with significant sequence diversity which are often part of multidomain proteins such as bacterial cellulose synthases. Members of the PilZ protein family regulate a variety of physiological processes such as polysaccharide biosynthesis, motility and virulence (Ryjenkov et al., 2006; Merighi et al., 2007; Pratt et al., 2007; McCarthy et al., 2008; Freedman et al., 2010; Pitzer et al., 2011). Upon cyclic di-GMP binding PilZ domains undergo dramatic conformational changes, which are the basis for differential protein–protein interactions and potential allosteric effects (Benach et al., 2007).

The mechanistic details of the mechanism how cyclic di-GMP regulates motility through interaction with the PilZ domain protein YcgR in E. coli and S. typhimurium are being unravelled. Upon binding of cyclic di-GMP YcgR can interact with flagella motor proteins, which slows down flagellar rotor speed and influences the frequency of the rotational switch thus affecting the velocity of the flagellar rotation and chemotaxis (Boehm et al., 2010; Fang and Gomelsky, 2010; Paul et al., 2010).

In GGDEF domain proteins, which are functional di-guanylate cyclases, cyclic di-GMP binding to the I-site leads to allosteric inhibition of the enzymatic activity thus tightly controlling cyclic di-GMP levels (Schirmer and Jenal, 2009). A functional I-site has also a specific role in the enzymatically deficient GGDEF domain protein PopA where it localizes the protein to the old cell pole in Caulobacter crescentus (Duerig et al., 2009). PopA localization subsequently leads to the degradation of the replication inhibitor CtrA at this site.

EAL domain proteins which have lost the cyclic di-GMP phosphodiesterase activity can still retain the substrate binding ability. Such a mechanism was demonstrated for the membrane-spanning GGDEF–EAL domain protein LapD (Newell et al., 2009). Cyclic di-GMP binding to the EAL domain leads to the recruitment of the periplasmic protease LapG to LapD which inhibits LapG functionality. The LapD/LapG pair is conserved in many bacterial species. Thus, proteolysis is emerging as another central physiological process regulated by the secondary messenger cyclic di-GMP (Duerig et al., 2009; Kumagai et al., 2010; Navarro et al., 2011).

Another level of complexity – conformational changes of the cyclic di-GMP molecule

So far, cyclic di-GMP binding sites cannot be predicted by bioinformatics. This is not only caused by a flexible build up of the cyclic di-GMP binding site by the cyclic di-GMP receptor proteins where very few amino acids are required to determine the binding specificity (Benach et al., 2007), but also by the flexibility in cyclic di-GMP conformation (Zhang et al., 2006; Wang et al., 2010). Monomeric cyclic di-GMP exist in an open and closed form (conformers) with respect to the orientation of the two guanine bases, which can be oriented to each other almost without energy hindrance. In addition, cyclic di-GMP is able to build up intermolecular interactions which leads to an equilibrium between the monomeric, dimeric, tetrameric and octameric form in solution (Zhang et al., 2006). This ability of cyclic di-GMP to form different conformers and aggregates adds to the complexity of signalling. First of all, aggregate formation of cyclic di-GMP molecules lowers the actual concentration of cyclic di-GMP available for receptor binding. Second, different cyclic di-GMP conformations are preferentially bound by different cyclic di-GMP receptor proteins. For example, the closed conformer is bound by most PilZ domain proteins (Wang et al., 2010). Dimer binding occurs at the I-site of GGDEF domain proteins (Chan et al., 2004) and also some PilZ domain proteins can bind dimeric cyclic di-GMP (Ko et al., 2010). As the presence of different physiologically available cations influences conformer/aggregate equilibrium (Zhang et al., 2006), differential binding of cyclic di-GMP polymorphs by receptors certainly contributes to the specificity of cyclic di-GMP signalling in bacterial cells.

Conclusions

Biofilm formation is probably the most ancient collective bacterial behaviour (Hall-Stoodley et al., 2004) and regulation of sessility versus motility may be the most ancient task of cyclic di-GMP signalling. However, the diversity of physiological processes and reactions regulated by cyclic di-GMP signalling and the sophisticated molecular mechanisms to provide a fine-tuned out-put response are intriguing. What is, however, even more intriguing is the specificity of the physiological output, namely the coordinated regulation of sessility and adherence versus motility and the virulence-persistence switch by the cyclic di-GMP secondary messenger molecule. Besides the regulatory mechanisms discussed in this review, additional mechanisms certainly also contribute to this specificity. Allosteric regulation of the enzymatic activity of the multi-domain cyclic di-GMP metabolizing proteins by intra- and extracellular signals (Barends et al., 2009; Rao et al., 2009; Schirmer and Jenal, 2009; Mills et al., 2011) is one of these mechanisms. In summary, the discovery and the subsequent analysis of cyclic di-GMP secondary signalling systems with its output specificity show that coordinated responses exist even in bacteria.

Acknowledgements

Work in the laboratory of the author was supported by the European Commission, the Swedish Research Council Natural Sciences, the Petrus and Augusta Hedlund Foundation, the Carl Trygger Foundation and the Karolinska Institutet.

Ancillary